Rheumatoid arthritis: an autoimmune disease that stopped destroying joints

Start here: what rheumatoid arthritis is

Rheumatoid arthritis is a systemic autoimmune disease in which the immune system attacks the synovium — the thin membrane lining the joints. The synovium becomes inflamed and thickened, and the resulting tissue erodes the cartilage and bone it sits against. Untreated, that erosion is permanent: joints deform, and function is lost for good.

The clinical picture follows from the biology. It is symmetrical, typically starting in the small joints of the hands and feet; it causes swelling and pain; and it produces prolonged morning stiffness — often an hour or more, which is the detail that most reliably separates it from ordinary joint pain. It is also systemic rather than confined to joints, with effects on the lungs, blood vessels, eyes and elsewhere ^{[1] [2]}.

It is not osteoarthritis, and the distinction is the whole story. Osteoarthritis is degenerative — cartilage wears down mechanically, typically in weight-bearing joints, typically asymmetrically, with stiffness that eases within minutes. Rheumatoid arthritis is immune: the damage is inflicted by the body's own inflammatory machinery, which is why it responds to drugs that suppress or redirect that machinery and why osteoarthritis does not. Confusing the two is the commonest public misunderstanding about this disease.

Why it matters — and why this review exists. Within living memory, RA meant progressive deformity and disability. It no longer does, for most patients treated early. That change came from three things arriving together: a way to measure disease activity numerically, a strategy of adjusting treatment until a target is reached, and a set of drugs that block specific immune pathways. The transformation is real and it is the reason the disease is worth explaining carefully ^{[1] [2] [3]}.

Why it opens the autoimmune domain here. RA is the flagship: the autoantibodies are measurable years before symptoms, the pathway targets are well characterised, and the therapeutic ladder from a cheap old tablet to precisely-aimed biologics is the template that other autoimmune diseases have followed ^{[4] [5]}.

Three pillars follow — measurements, medicines, and progress — with a model between the first two built on the trial that most clearly separates feeling better from not being destroyed.

Pillar 1: measurements and diagnosis

Two autoantibodies, and what they know before you do

Rheumatoid factor (RF) is an antibody directed against the constant region of other antibodies. It is the older test, and it is sensitive but not specific — it appears in other conditions and in some healthy people.

Anti-citrullinated protein antibodies (ACPA), measured clinically as anti-CCP, are far more specific for RA. Citrullination is a normal post-translational modification; in RA the immune system starts treating citrullinated proteins as foreign.

The remarkable finding is when they appear. In a study of blood donors who later developed RA, with a median of 13 stored serum samples each and the earliest collected a median of 7.5 years before symptoms, 49% were positive for IgM-RF and/or anti-CCP on at least one occasion before any symptoms, at a median of 4.5 years before symptom onset. Among 2,138 control samples, only 1.1% were RF-positive and 0.6% anti-CCP-positive ^[6]. Antibodies against cyclic citrullinated peptide predict who will develop RA ^{[7] [8]}, the autoantibody response broadens in the pre-clinical phase in a way that itself predicts progression ^[9], and cytokine and chemokine up-regulation also predates onset ^[10].

So the disease has a pre-clinical phase, measurable, lasting years. That fact is the basis of the interception research in Pillar 3, and it is unusual: most diseases do not announce themselves half a decade in advance.

Inflammatory markers and classification

CRP and ESR are non-specific measures of systemic inflammation. They cannot diagnose RA on their own, but they quantify how much inflammation is present and they feed into the composite scores below.

Diagnosis in practice uses the 2010 ACR/EULAR classification criteria, which score joint involvement, serology (RF and ACPA), acute-phase reactants and symptom duration ^{[11] [12]}. These were designed deliberately to identify disease earlier than the criteria they replaced, because — as Pillar 2 shows — early treatment is what preserves joints.

DAS28: the number that changed the disease

The single most consequential measurement in RA is a composite. DAS28 combines a count of tender joints and swollen joints out of 28, an inflammatory marker, and the patient's own global assessment, into one number. Thresholds define high, moderate and low disease activity, and DAS28 below 2.6 is conventionally called remission ^[13].

Why this matters more than it sounds: a single number that a clinician and patient can both see, at every visit, makes it possible to aim. Alternative composites exist — the SDAI and Boolean-based definitions apply stricter criteria for remission — and the definition chosen materially changes how many patients count as being in remission ^[14].

Imaging: seeing the damage

Plain radiographs show erosions and joint-space narrowing, scored using systems such as the modified total Sharp score — the outcome that appears in the model below ^[13]. Ultrasound and MRI detect synovitis and erosions earlier and more sensitively than radiography.

But more sensitive is not automatically better, and the field tested this honestly. The ARCTIC trial randomised 238 early-RA patients to a treatment strategy targeting clinical and imaging remission versus one targeting clinical remission alone, with the same escalation protocol and 13 visits over two years. The primary endpoint was reached by 22% versus 19% — mean difference 3.3%, 95% CI −7.1 to 13.7 — with similar disease activity, function and joint damage. Its conclusion was blunt: the systematic use of ultrasound in follow-up "is not justified on the basis of the ARCTIC results" ^[15]. A better picture of inflammation did not translate into better outcomes once treatment was already being adjusted properly.

Centerpiece: a simple simulatable model of what treatment protects

The measurements above exist to answer one question: is this patient's disease being controlled well enough to stop the joint destruction? The trial that separates that question from "does the patient feel better" most cleanly is PREMIER.

PREMIER randomised 799 patients with early, aggressive, methotrexate-naive RA to adalimumab plus methotrexate, methotrexate alone, or adalimumab alone, double-blind, for two years, with co-primary endpoints of symptomatic response and radiographic progression ^[13]. Because it reports both axes for all three arms at both years, it can be read as two different curves for the same three strategies.

What the model explains. Four things.

First, that symptoms and damage are different axes. On symptoms the three arms are close: 62% versus 46% versus 41% reaching ACR50 — a 1.35-fold spread. On joint destruction they are not close at all: 1.9 versus 5.5 versus 10.4 Sharp units at two years — a 5.5-fold spread. A patient on methotrexate alone reporting a good response was, on average, still accruing damage at nearly eight times the rate of a patient on combination therapy. This is why RA is managed by measurement rather than by how the patient says they feel.

Second, why "early" is doing so much work. Two years of combination therapy produced less structural damage than one year of methotrexate alone. Damage accrued during a delay is not recovered later — the curves never come back down. That asymmetry is the entire argument for treating early and escalating fast, and it is why the classification criteria were rewritten to catch people sooner ^[11].

Third, why the two monotherapies look so similar and the combination does not. Methotrexate alone and adalimumab alone differ by 5 percentage points on ACR50; the combination is 16 points above the better of them. The two drugs are not redundant — they suppress different things, and the trial's own conclusion is that combination was superior to either alone on signs and symptoms, radiographic progression and remission ^[13]. Combination therapy in early RA was tested in other strategy designs too, including attenuated versions of intensive combination regimens ^[16].

Fourth, why remission became a realistic target at all. Forty-nine per cent of the combination arm reached DAS28 remission at two years ^[13]. In the pre-biologic era that was not an expected outcome, and its arrival is what made "treat to target" a coherent strategy rather than an aspiration.

What the model deliberately does not do. It is one trial in early, aggressive, seropositive-enriched disease; established RA behaves differently and the monotherapy arms are not current practice. The Sharp score is ordinal, so the ratios above describe radiographic scores rather than proportional physical joint loss. Two timepoints cannot establish the shape of accrual within a year. And it says nothing about cost, access or the long-term safety that Pillar 2 has to weigh.

Pillar 2: medicines — a ladder with rungs of different kinds

Methotrexate: the anchor

Almost every RA treatment path starts with methotrexate, a weekly oral (or injected) tablet that has been the anchor conventional DMARD for decades and remains so. It is a folate antagonist — its classical target is dihydrofolate reductase — but its anti-inflammatory effect at the low weekly doses used in RA is not simply antiproliferative. Increased adenosine release at inflamed sites, which diminishes leukocyte accumulation, was identified as a mechanism decades ago ^[17], and the drug's pharmacology continues to be re-examined ^[18]. It requires folate supplementation and monitoring of blood counts and liver function.

Two things keep it first-line. It works: 46% reached ACR50 on methotrexate alone in PREMIER ^[13]. And it makes the biologics work better — every combination arm in the trial literature pairs a biologic with methotrexate rather than replacing it.

Biologics, by mechanism

When methotrexate is insufficient, the next rung targets a specific immune pathway. Four mechanisms are established, and they are genuinely different — which matters, because a patient failing one can respond to another.

TNF inhibition blocks tumour necrosis factor α, a central inflammatory cytokine. This was the first class to arrive and it reset expectations for the disease; adalimumab is the agent in the model above ^[13]. The class is now large, its place is set out in the EULAR management recommendations and their successive updates ^{[19] [20] [21] [22]}, and the comparative efficacy evidence behind those recommendations has been systematically assembled ^[23]. Its overall role has been reviewed repeatedly ^{[3] [24]}.

IL-6 pathway blockade — tocilizumab — targets interleukin-6 receptor signalling, established in RA by the OPTION trial ^[25] against a cytokine whose role across inflammation and disease is well characterised ^[26]. IL-6 drives the acute-phase response, which has a practical consequence worth knowing: blocking it drops CRP and ESR profoundly, so the inflammatory markers inside DAS28 fall partly for pharmacological rather than purely disease-related reasons — and DAS28 itself has separate validated variants depending on which marker is used ^[27].

B-cell depletion — rituximab, an anti-CD20 antibody — removes the B lymphocytes that produce autoantibodies and present antigen. Its efficacy in RA was established in a randomised trial of B-cell-targeted therapy ^[28], and it is notable for working in a disease long modelled as T-cell-driven.

T-cell co-stimulation blockade — abatacept, a CTLA-4–immunoglobulin fusion protein — intercepts the second signal that T cells need to become fully activated, acting further upstream than the cytokine blockers. It was shown to work in patients whose disease was refractory to TNF inhibition ^[29] — evidence that failing one mechanism does not predict failing another — and was compared head-to-head against infliximab in ATTEST ^[30]. The underlying CD28 co-stimulation biology has been reviewed in its own right ^[31].

Targeted synthetic DMARDs: the JAK inhibitors, and an honest safety caveat

The newest rung is oral rather than injected. JAK inhibitors — tofacitinib and successors — block Janus kinases inside the cell, interrupting signalling downstream of many cytokine receptors at once. The relevant enzymes are JAK1, JAK2, JAK3 and TYK2, and which of them a drug inhibits, and how selectively, is the axis along which the class differentiates itself.

Their efficacy is not in question. Their safety required a dedicated trial, and the result deserves to be stated plainly rather than buried. ORAL Surveillance randomised 4,362 patients with active RA despite methotrexate, aged 50 or over with at least one additional cardiovascular risk factor, to tofacitinib 5 mg twice daily, 10 mg twice daily, or a TNF inhibitor, and followed them a median of four years. Incidences of major adverse cardiovascular events and cancer were higher with the combined tofacitinib doses (3.4% and 4.2%) than with a TNF inhibitor (2.5% and 2.9%); hazard ratios were 1.33 (95% CI 0.91–1.94) for MACE and 1.48 (1.04–2.09) for cancers, and non-inferiority was not shown. Opportunistic infections, herpes zoster and non-melanoma skin cancer were also more frequent. Efficacy was similar across all three groups ^[32].

That trial enrolled a deliberately risk-enriched population, and how far its findings generalise has been actively argued ^[33], with observational and open-label comparisons examining cardiovascular ^[34] and malignancy ^[35] risk against TNF inhibitors. The reasonable reading is that JAK inhibitors are effective drugs whose risk-benefit balance depends on the individual patient's cardiovascular and cancer risk — not that they are unsafe, and not that the signal can be dismissed.

Treat-to-target — and how good the evidence actually is

The strategy tying all of this together is treat-to-target: pick a target (remission, or low disease activity), measure disease activity at regular intervals, and escalate therapy until the target is reached. Tight control has been examined for efficacy and feasibility ^[36] and followed over ten years in recent-onset disease ^[37], and the approach is embedded in management recommendations.

It would be easy to present that as settled. It is not, quite. A systematic review of 16 clinical effectiveness studies found that they differed in target, protocol and visit frequency, that heterogeneity precluded meta-analysis, and that "for several outcomes, mixed results or evidence of no difference between TTT and conventional care was found" — with favourable remission rates in early disease in two studies, one not statistically significant ^[38]. ARCTIC, above, found that intensifying the target further with imaging added nothing ^[15].

The honest position: measuring disease activity and escalating on it is the right default and is what the modern outcomes were achieved under — but the incremental benefit of formal protocolised treat-to-target over attentive conventional care is smaller and less consistent than the strength of the recommendation implies.

Pillar 3: progress

Intercepting the disease before it starts

The pre-clinical window in Pillar 1 — autoantibodies a median of 4.5 years before symptoms ^[6], with epitope spreading ^[9] and cytokine changes ^[10] along the way — raises an obvious question: can RA be prevented in people who are antibody-positive but not yet arthritic? Trials of interception in at-risk individuals are the field's most ambitious current direction. The scientific case is strong; the ethical bar is high, because the intervention is immunosuppression in someone who is currently well and who may never develop the disease.

Biosimilars and access

The biologics transformed outcomes and were, for years, extremely expensive. Biosimilars — follow-on versions of off-patent biologics — have brought prices down substantially ^[39], and their efficacy and drug survival have been tracked in inflammatory disease ^[40]. This is a genuine advance in a disease where the gap between best available and typically received care has been largely economic rather than scientific. Immunogenicity — patients forming antibodies against the biologic itself — remains a consideration across the class ^[41], as do its adverse effects, assembled in a network meta-analysis and Cochrane overview ^[42].

Precision selection

The remaining clinical problem is that no test reliably predicts which mechanism will work for which patient. Choice between a TNF inhibitor, IL-6 blockade, B-cell depletion, co-stimulation blockade and a JAK inhibitor is still substantially trial-and-error, and each failed trial costs months of disease activity — months in which, as the model shows, joints are being lost. Progress in understanding RA aetiology ^[5] and autoimmune pathogenesis generally ^[4] is the route to fixing that, and it has not been fixed yet.